INDEX

Explanations

words related to concepts of controversy or conflict

oai_token-act-pair · gpt-3.5-turbo

references to confidentiality or related themes

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ï¸ı

-0.71

senal

-0.67

DAY

-0.67

Ô

-0.67

BALL

-0.66

OHN

-0.64

NING

-0.63

hyde

-0.62

Â¯Â¯Â¯Â¯

-0.61

 Roof

-0.60

POSITIVE LOGITS

ederation

1.34

essional

1.27

luence

1.27

usions

1.16

eder

1.16

erences

1.13

idences

1.07

essor

1.05

ederal

1.04

eren

1.00

Activations Density 0.014%

words related to concepts of controversy or conflict

references to confidentiality or related themes

No Comments

No Known Activations